Getting Started with Intel AI Devcloud.
Scenario 1- It was a few weeks back; I was running my deep learning model for image classification. It contained almost ten epochs. It used to take nearly 20 minutes for each epoch to run. I used to get errors randomly in-between process, so I had to start from the beginning. Sometimes power issues due to the battery used to disrupt the process. It used to take me days to complete one deep learning model. It was a cumbersome task.
Scenario 2- I was introduced to this fantastic technology — Intel® AI DevCloud. It is a cluster of Intel® Xeon® Scalable Processors that will assist you with your machine learning and in-depth learning training and inference compute needs. Now nothing stops me from running my models. Guess what? I can run the same models within minutes. Credit goes to Intel Devcloud.
What Is Intel Devcloud?
The DevCloud is a cluster of Intel® Xeon® Scalable Processors that will assist you with your machine learning and deep learning training and inference compute needs. It provides access to precompiled software optimized for Intel® architecture on Intel® Xeon® Scalable Processors. It includes:
- Intel® Software Optimization for neon™
- TensorFlow* on Intel® Architecture
- Intel® Optimizations for MXNet*
- Intel® Distribution for Caffe*
- Intel® Distribution for Python* 2.7 and 3.6 including NumPy, SciPy, pandas, scikit-learn, Jupyter, matplotlib, and mpi4py
- Intel® Software Optimization for Keras*
- Intel® Software Optimization for Theano*
- Intel Nervana AI platforms and technologies that become available in the future
When you gain access to the DevCloud, you will log into a Linux-based head node of a batch farm. There you can stage your code and data, compile, and submit jobs to a queue. Once the queued job completes, your results will be in your home folder.
- Jobs are scheduled on Intel® Xeon® Scalable Processors.
- Each processor has 24 cores with 2-way hyper-threading.
- Each processor has access to 96 GB of on-platform RAM (DDR4
- Only one job will run on any processor at a time.
- You will get 200 GB of file storage quota.
- Your home directory is not visible to other users.
- Once your access period expires, your home directory on the cluster will be deleted.
Who can sign up to use the Intel® AI DevCloud?
Developers, data scientists, professors, students, start-ups and others who are members of Intel® AI Academy are eligible to request access.
How do I become a member of the Intel® AI Academy?
You can join the AI Academy by requesting access to the DevCloud here, or you can become a member by registering here.
Next Steps-
After you signup successfully for AI cloud, you will receive your credentials to your registered mail. Once you receive mail you have two approaches to access devcloud. Do not lose this email. You can directly login using specific email presented in your mail-id. If not, click here. After you click on the link, you can see information on-page to get started. You will see content, as shown in the image below.
1.Using Jupyter notebook
Your Jupyter Notebook instance runs on one of the compute nodes. When you execute the Notebook cells, they run on an Intel Xeon Scalable processor. There may be other people connecting to your compute node, which reduces the available compute power. However, you can reserve a full node for your job and even use multiple nodes at once by submitting scripts to the job queue. For more information, connect to a Jupyter Notebook instance and open the Welcome.ipynb file in your home directory.
If you wish to install libraries from outside, , you can use standard PIP or Conda commands to install. Use terminal, as mentioned in below image to install libraries.
In terminal type, pip install <Package name> to install the package on devcloud.
After you get connected, you can use any of the kernels given or install new kernels as required to run your model. I used to upload necessary data to the cloud and run it from Jupyter notebook.
2. Using Putty
click here to learn how to access devcloud using putty.
Eight-word summary: do not run jobs on the login node.
When you log in, you will find yourself on the host login-1, which is your login node. This node is intended only for code development and compilation, but NOT for computation. That is because it does not have much compute power, and, additionally, there are limitations on CPU time and RAM usage on the login node; your workload will be killed if it exceeds the limit. To run computational workloads on powerful compute nodes, you must submit a job through the Torque job queue. See the next section for a sample job script.
You can find more detailed information about jobs here.
You can also go through mentioned links below to for further details about devcloud.
Useful Links-
- Link to Webinar on Introduction to devcloud-
https://software.intel.com/en-us/videos/get-started-with-the-intel-ai-devcloud
2. https://forums.intel.com/s/question/0D50P0000490RjGSAU/devcloud-usage-basics
3. https://software.intel.com/en-us/
References-
